摘要
In this paper, a novel reconstruction technique based on level set method and algebraic reconstruction technique is proposed for multiphase flow computed tomography (CT) system. The curvature-driven noise reduction method is inserted into the conventional iteration procedure of algebraic reconstruction technique to improve the image quality and convergence speed with limited projection data. By evolving the image as a set of iso-intensity contours after each updation, the sufficient number of iterations for acceptable results is reduced by 80%-90%, while the image quality is enhanced obviously. Quantitative evaluation of image quality is given by using both relative image error and correlation coefficient. The resultant images can be utilized to detect flow regimes for monitoring industrial multiphase flow. Laboratory results demonstrate the feasibility of the proposed method. Phantoms of four typical flow regimes can be reconstructed from few-view projection data efficiently, and the corresponding image errors and correlation coefficients are acceptable for the cases tested in this paper.
In this paper, a novel reconstruction technique based on level set method and algebraic reconstruction technique is proposed for multiphase flow computed tomography （CT） system. The curvature-driven noise reduction method is inserted into the conventional iteration procedure of algebraic reconstruction technique to improve the image quality and convergence speed with limited projection data. By evolving the image as a set of iso-intensity contours after each updation, the sufficient number of iterations for acceptable results is reduced by 80%--90%, while the image quality is enhanced obviously. Quantitative evaluation of image quality is given by using both relative image error and correlation coefficient. The resultant images can be utilized to detect flow regimes for monitoring industrial multiphase flow. Laboratory results demonstrate the feasibility of the proposed method. Phantoms of four typical flow regimes can be reconstructed from few-view projection data efficiently, and the corresponding image errors and correlation coefficients are acceptable for the cases tested in this paper.
基金
Supported by National Natural Science Foundation of China (No.60820106002,No.60532020)
Natural Science Foundation of Tianjin(No.11JCYBJC06900)
